Image Convolution
convolution application
Multimedia \Special effects\ photo enhancement\ computer image analysis
convolution calculate step
1. position the center of the kernel at the first pixed of an image and flip the kernel(选择180度)
2. Multiple and sum
3. slide to next pixel
边缘点处理:1. 丢弃边缘 2.边缘插值
小于0为0 ,大于255为255
Image Gradient and Edge detection
An edge is s set of connected pixels the lie on the boundary between two regions
How it comes
Different colors, brightness, textures, material, tissues
Different normal direction of surface
Different illuminance
Edge Category
Gradient
A point is defined as an edge point if its 2-D first or second-order derivativeis greater than a special threshold
gradient can also be interpreted as a complex number with its magnitude determined by:
and the direction determined by:
Roberts
Prewitt
Sobel
Image Noise Reduction
coherence -- incoherence
Gaussian\Median Filter\ open\ close
Gaussian reduction (更好地保持图像原来的属性) standard deviation越大,效果越强,越模糊
Mean Filter(kernel尺寸越大,效果越强,越模糊)
Median Filter(排序去中间值)